What problem does it solve? Turning a research narrative report into a submission-ready academic paper requires coordinating outlining, figure generation, LaTeX writing, compilation, and iterative review — a fragmented, error-prone process when done manually. ## Core Features & Use Cases - End-to-End Pipeline: Chains paper-plan, paper-figure, paper-write, paper-compile, and an auto-improvement loop into one automated workflow producing a polished paper/ directory with LaTeX source and PDF. - Multi-Layer Quality Gates: Includes adversarial acceptance-contract negotiation, proof verification, claim audits against raw result files, citation audits, and kill-argument review for theory-heavy papers. - Flexible Illustration Modes: Generates architecture and method figures via deterministic JSON-to-SVG, Gemini, Codex image bridge, or Mermaid depending on figure type and available API keys. - Use Case: A researcher with a NARRATIVE_REPORT.md containing claims and experiment results invokes the pipeline to produce an ICLR-formatted paper with auto-generated plots, verified citations, and two rounds of GPT-5.5 review-driven polishing. ## Quick Start Run the paper writing pipeline on my NARRATIVE_REPORT.md targeting ICLR with default settings.
